43

Having spent many hours looking through the Atlas of Finite Simple Groups while in Grad school, I recall being rather intrigued by the fact that among the sporadic groups, only one (McLaughlin as I recall, and only 2 out of its 24 irreps are quaternionic) has any irreps of quaternionic type. On the other hand, to my recollection several members of infinite families (such as those arising from the symplectic groups) as well as certain covers of the sporadics have quaternionic irreps. As I do not currently have access to the Atlas, I can't really list a bunch of examples, but if you have access to a copy you can go look them up.

Question: Is there a 'natural' reason that quaternionic representations and simple groups (in particular sporadics) like to avoid one another? Specifically, is there something intrinsic about preserving a symplectic form which implies that the corresponding automorphism group "should" have a normal subgroup (because of something trivial like symmetry considerations)?

ARupinski
  • 5,181
  • 2
    My guess is something like "it's hard for large 2-groups to have symplectic representations, so any simple group with one must have low 2-rank." – Ben Webster Feb 08 '11 at 19:15
  • This may well be a hard question, but first I'd wonder whether your Atlas observations about groups of Lie type hold up more generally? The Atlas covers all sporadics but only a "few" of the others. (By the way, for quaternionic representations via Frobenius-Schur indicators, Serre's Springer text has a nice treatment in 13.2 even though those terms aren't easy to locate in the index.) – Jim Humphreys Feb 08 '11 at 19:36
  • @Jim Although I don't offhand know of about whether the Lie-type groups do indeed continue to display the same sort of behavior, my guess would be yes based on the fact that these series tend to have lots of properties in common from one member to the next. On the other hand, it is conceivable that the existence of quaternionic irreps could depend on the characteristic of the underlying field (So for example maybe $C_n(p)$ has quaternionic irreps only for certain $p$). If this is the case surely someone must have studied which primes allow quaternionic irreps and which don't. – ARupinski Feb 08 '11 at 20:05
  • 1
    It looks like a lot of unitary groups and a lot of twisted groups in general have quaternionic irreducible ordinary representations. It might not be rare, and doesn't seem like it requires low 2-rank. – Jack Schmidt Feb 09 '11 at 00:05
  • 7
    Srinivasan–Vinroot have a preprint showing unitary groups of even degree have a ton of quaternionic irreps, and gives a reason why. http://www.math.wm.edu/~vinroot/SemiSympChars-REV.pdf – Jack Schmidt Feb 09 '11 at 00:09
  • 4
    @Jack: From your comments and the sharp counts of quaternionic representations in the S-V paper, it seems plausible that if one starts with symplectic forms and tries to determine which finite simple groups can leave such forms invariant, one would eventually determine some families of representations of classical groups and two sporadic representations (of the McLaughlin Group). Thus it looks like my question would be equivalent to a classification problem and hence likely difficult. – ARupinski Feb 09 '11 at 07:42
  • 1
    That is an interesting question, which I've thought about in the context of Lie groups. If G is a compact connected Lie group then the real/quaternionic indicator (equal to the Frobenius-Schur indicator) of a representation is its central character evaluated at a certain element. So (for example) if G is adjoint then every irreducible self-dual representation is real, and in general "at most half" of representations are quaternionic. This is more of an observation than an explanation. See http://arxiv.org/abs/1203.1901 – Jeffrey Adams Jun 06 '16 at 20:41
  • My guess would be "It's hard to have a subgroup of a group that involves a lot of quaternions without hitting $-1$ somehow", especially since, for instance, the double cover of the Hall-Janko group is the symmetry group of a quaternionic (icosian) structure on the Leech lattice – W. Cadegan-Schlieper Jul 29 '17 at 17:59
  • For any finite $G$ group of even order, the standard formula telling us that $\sum_{\chi \in {\rm Irr}(G)} \nu(\chi) \chi(1)$ is the number of solutions of $x^{2} = 1$ in $G$, indicates that the sum of the degrees of (real-valued complex) irreducible characters with FS-indicator $+1$ is much larger than the sum of the degrees of the irreducible characters with indicator $-1$. – Geoff Robinson Jul 27 '23 at 12:22

0 Answers0